Both the model and parameter uncertainties should be taken into consideration for evaluation of soil liquefaction potential during earthquake. The paper aims to develop a new method to quantify the model uncertainty of the simplified method. The proposed method is constructed by the random simulation technique. The CPT-based liquefied/non-liquefied case histories are collected from documents to determine the Roberson and Wride(1998) model (RW) uncertainty for example. The results show that the RW model uncertainty can be defined by a c1 parameter with an expected value is 1.16 and a coefficient of variance value is 0.12 for all case histories. Finally, results comparison is presented for the validity of the proposed method.
In the simplified methods of soil liquefaction by earthquake, one mathematics function is often used to develop a limit state that separates liquefied case form non-liquefied case. The limit state of simplified method is an empirical model through a great quantity of history cases and test data. If a mathematics function can be used to express the idealized limit state of simplified method; it is referred herein as ILS. The ILS is closest to the real limit state; it is referred herein as RLS. Take the simplified model by Roberson and Wride (1998) as an example; it is referred herein as RW model. Although the several calibrations to the original RW model were addressed because of the increasing case histories, the original mathematics function still was followed (Hwang and Yang, 2001; Youd and Idriss, 2001). One often uses enough case histories to determine the empirical model shown as Figure 1. In Figure 1, the black dot and hollow dots mean the liquefied and non-liquefied cases respectively. The black line marked by C represents the state of the highest success rate, in which only three cases are misclassified.
